The Gold Standard of Ecology | Nest Survival Models, Exposure Days and Visit Timing

1. Define the daily survival question

Back to contents

Daily survival rate is the probability that an active nest survives one day under the model. Overall success across a nesting period is the product of daily survival across those days. If daily survival is constant at D for 28 days, the period probability is D to the power of 28. Small daily differences can therefore create large seasonal differences.

State the biological interval: laying to fledging, incubation only, or another stage. Define success and failure for the species. A fledged nest, an empty nest with no sign, and a nest abandoned after disturbance may require different evidence rules.

2. Use exposure rather than the found-nest denominator

Back to contents

Nests found late have had to survive long enough to be discovered, so a raw success proportion is biased upward. Exposure methods count the time each monitored nest was known to be at risk. A nest contributes from first active observation until failure, success, censoring or the chosen end.

The risk interval between visits is known even when the precise event time is not. Models use its length directly. Record dates and times consistently; a three-day interval is not equivalent to one daily check repeated three times. Exposure is a design variable, not just a spreadsheet total.

3. Build one auditable history per nest

Back to contents

For every nest, retain discovery date, last active date, first failed or successful date, interval lengths, stage, observer, visit reason and fate evidence. Mark nests that were never rechecked or whose outcome is uncertain. Do not back-calculate exact failure times unless a justified interval rule or sensor provides them.

Renesting and shared parents need identity rules. Locations may require protection, but anonymised histories and analysis code can remain auditable. The final encounter history should be reproducible from field notes rather than assembled from memory after the season.

4. Understand the model’s likelihood

Back to contents

A nest that survives a seven-day interval contributes a survival probability raised to seven; a nest known to fail over that interval contributes survival through earlier days and failure by the end under the chosen formulation. Logistic-exposure and Program MARK nest-survival approaches can include individual and time-varying covariates while accommodating unequal intervals.

Independence, correct fate and suitable functional form still matter. Several nests in one colony or monitored by the same observer may share predators and disturbance. Use clustering or hierarchical structure when the design and sample support it.

5. Treat discovery and visits as observation processes

Back to contents

Nest visibility changes with vegetation, stage, adult behaviour and observer effort. Easily found nests may have different survival. Visits can leave scent trails, flush adults, reveal nests to predators or, conversely, deter predators. Check schedules may intensify near expected fledging, making interval length depend on perceived fate.

Record search effort and reasons for visits. Randomised or standardised visit schedules can separate observation from biology. Compare early and late-found nests and test short-term hazard after visits where ethical design permits. A flexible model cannot repair unrecorded disturbance.

6. Worked example: exposure days change the answer

Back to contents

In an invented study, 30 nests contribute 420 known exposure days and six confirmed failures. A simple constant-rate illustration gives daily survival of 414 divided by 420, about 0.9857. Over a 28-day nesting period, 0.9857 to the power of 28 is about 0.67. Reporting only 24 of 30 nests not known to fail would suggest 0.80 and ignore unequal time at risk.

The calculation is illustrative; fitted nest-survival models use interval histories and uncertainty more carefully. The lesson is that daily risk compounds. A rate that looks close to one does not imply that most nests complete a long nesting cycle.

7. Handle uncertain fate and interval timing

Back to contents

An empty nest may have fledged, been predated, been abandoned or become unobservable. Use species-specific signs, cameras or narrow revisit windows where justified. Blindly coding every empty nest as success or failure can dominate the estimate. Censor truly unresolved outcomes and test assumptions about them.

Exact failure time may be interval-censored. Midpoint assignment is convenient but can distort relationships with weather or nest age. Models that use visit intervals avoid false precision. If a storm occurs inside a long interval, the data may not identify whether it preceded or followed failure.

8. Changed-context transfer check: sea-turtle nests

Back to contents

Move from bird nests checked every few days to sea-turtle nests monitored from laying to emergence. Exposure logic transfers, but failure definitions, observation frequency and stages differ. Flooding, excavation, partial hatching and emergence counts add multistate outcomes. Human tracks may affect disturbance differently.

The transferable principle is to align risk intervals and fate evidence with the organism. A bird-focused 28-day constant-rate summary cannot be copied directly. The target period, observation process and competing failure modes must be rebuilt for the new life history.

9. Know the limits and common misconceptions

Back to contents

Daily survival is not the same as nesting success unless it is compounded over a defined period. A higher estimate can result from finding nests later, not safer habitat. More visits can improve fate resolution while changing fate. Model selection weights do not prove causation for habitat covariates.

Small numbers of failures limit complex models. Date, nest age and weather may be correlated, causing unstable coefficients. Apparent precision can ignore colony clustering and uncertain fate. Report the data support for each effect and avoid translating association directly into management action.

10. Audit field design and model together

Back to contents

Publish search routes, discovery effort, visit schedules, disturbance safeguards, fate criteria, exposure construction and exclusions. Report nests, exposure days, failures, successes, censoring and uncertain outcomes by stage and group. Preserve interval histories and covariate sources.

Provide candidate-model rationale, link function, variance treatment and predictions on an interpretable scale. Refit under plausible fate reclassifications, alternative period lengths and visit-effect windows. Plot daily and period survival with uncertainty. The reader should see how field observation becomes each likelihood contribution.

11. Frequently asked questions

Back to contents

Why not divide successful nests by nests found? Late-found nests are conditioned on earlier survival and observation time differs. What is an exposure day? One day during which a monitored nest is at risk under the defined history. Does 0.98 daily survival mean 98% succeed? No; survival compounds across many days.

Can irregular checks be used? Yes, if interval lengths and fates are recorded. Should uncertain empty nests be failures? Not automatically. Do visits cause predation? They can alter risk in some settings, so design and sensitivity checks matter.

12. The gold standard joins exposure to observation

Back to contents

Nest survival models earn their value by respecting when each nest enters observation and how long it remains at risk. That improvement is lost when discovery bias, visit timing or uncertain fate is hidden. The daily rate must be translated into the full biological period readers care about.

The gold standard preserves interval histories, models observation-related risks, limits complexity to supported data and tests fate rules. It produces a survival estimate that is traceable from field visit to ecological claim—and clear enough to guide monitoring without overstating certainty.